CMR#

Tools for implementing variants of the context maintenance and retrieval model (CMR).

The cmr module provides tools for implementing various CMR model variants using the CMR class. Rather than implementing one specific version of CMR, CMR is designed to be highly configurable via changes to the CMRParameters class, which supports various ways of defining network weights and parameter values.

Model variants may be evaluated against observed data by calculating likelihood, or fitted to data using a parameter search. They can also be used to generate simulated data.

Model variants are defined by CMRParameters objects, which may be saved to JSON files, and patterns, which are stored in HDF5 format. The patterns define representations associated with individual items. The patterns may be orthonormal, as in the original implementations of TCM and CMR, or they may be overlapping, as in DCMR. CMRParameters specify how patterns are used to determine weights in the network, among other model settings.

Model framework#

CMR()

Context Maintenance and Retrieval model.

CMR.likelihood(data, group_param[, ...])

Log likelihood summed over all subjects.

CMR.fit_indiv(data, param_def[, patterns, ...])

Fit parameters to individual subjects.

CMR.generate(data, group_param[, ...])

Generate simulated data for all subjects.

CMR.record(data, group_param[, subj_param, ...])

Record model states during a simulation.

CMR.parameter_sweep(data, group_param, ...)

Simulate data with varying parameters.

CMR.parameter_recovery(data, n_sample, param_def)

Run multiple iterations of parameter recovery.

Model configuration#

save_patterns(h5_file, items, **kwargs)

Write patterns and similarity matrices to hdf5.

load_patterns(h5_file[, features])

Load weights from an hdf5 file.

read_config(json_file)

Read model configuration from a JSON file.

config_loc_cmr(n_item)

Configure a localist CMR network.

Model parameters#

CMRParameters()

Configuration of CMR model parameters.

CMRParameters.set_fixed(*args, **kwargs)

Set fixed parameter values.

CMRParameters.set_free(*args, **kwargs)

Set free parameter ranges.

CMRParameters.set_dependent(*args, **kwargs)

Set dependent parameters in terms of other parameters.

CMRParameters.eval_dependent(param)

Evaluate dependent parameters based on input parameters.

CMRParameters.set_dynamic(trial_type, scope, ...)

Set dynamic parameters in terms of parameters and data.

CMRParameters.eval_dynamic(param[, study, ...])

Evaluate dynamic parameters based on data fields.

CMRParameters.set_sublayers(*args, **kwargs)

Set layers and sublayers of a network.

CMRParameters.set_weights(connect, regions)

Set weights on model patterns.

CMRParameters.set_sublayer_param(layer, ...)

Set sublayer parameters.